FrequencyCT: Frequency domain pseudo-label generation for self-supervised low-dose CT denoising

2026-05-11

AI Summary

This work addresses the challenge of effectively modeling complex noise correlations in the projection domain of low-dose CT imaging. The authors propose a novel frequency-domain zero-shot self-supervised denoising method that constructs self-supervised pseudo-labels directly in the frequency domain for the first time. By anchoring low-frequency components and preserving phase information through amplitude modulation, the approach retains essential structural details, while high-frequency masking perturbations enhance diversity. A projection-domain sample truncation strategy is further introduced to stabilize training gradients. This framework overcomes the limitations of conventional spatial-domain self-supervision by effectively decoupling noise from clean signals. Extensive experiments on multiple public and real-world clinical datasets demonstrate that the proposed method significantly outperforms existing state-of-the-art techniques, highlighting its strong potential for clinical deployment.

Abstract

Despite extensive research on computed tomography (CT) denoising, few studies exploit projection-domain data characteristics to mitigate noise correlation. To address this, this work proposes FrequencyCT, the first zero-shot self-supervised method for pseudo-label generation in the frequency domain for low-dose CT denoising. Leveraging the characteristic of the frequency domain that largely isolates noise from clean signals, a regional low-frequency anchoring technique is proposed. Phase-preserving amplitude modulation and mask perturbation in the high-frequency region generate pseudo-label data for self-supervision. The fluctuating noise variance in the projection domain prompts truncation of the generated samples to stabilize the network's optimization gradient. Evaluation results on multiple public and real-world datasets confirm the clinical application potential of this research, which will have a revolutionary impact on the field of denoising. The code can be obtained from https://github.com/yqx7150/FrequencyCT.

Problem

• low-dose CT denoising
• noise correlation
• projection-domain data
• self-supervised learning
• pseudo-label generation

Innovation

• frequency domain
• pseudo-label generation
• self-supervised learning
• low-dose CT denoising
• zero-shot